Merge pull request #65 from Fersoil/master

add glaze presentation
MI2DataLab · Jan 7, 2024 · a144f75 · a144f75
2 parents f02ca18 + bbea0bd
commit a144f75
Show file tree

Hide file tree

Showing 3 changed files with 6 additions and 1 deletion.
diff --git a/2023/2023_12_18_glaze_protecting_artists_from_style_mimicry/README.md b/2023/2023_12_18_glaze_protecting_artists_from_style_mimicry/README.md
@@ -0,0 +1,5 @@
+# Glaze: Protecting Artists from Style Mimicry by Text-to-Image Models
+
+Generative AI models have drastically changed how we create and consume images and art. Models like MidJourney and Stable Diffusion have been trained on large datasets like LAION-5B containing scraped private or copyrighted images. As a result, many artists were placed before deciding whether to upload their art online and risk their artwork being illegally used in models' training or losing a vital way of promoting their artwork. An even more significant threat for artists is "style mimicry" - a specific attack in which a model is fine-tuned on artists' images. The final tuned model can produce images in the personal style of the target artists, which usually takes years for the artist to achieve. During the following Monday's presentation, I will briefly recap how Stable Diffusion works and present a tool that helps artists protect their style. Glaze is a solution that allows disrupting the image enough; in a way, the specific artist's style can't be extracted from the data.
+
+The presentation will be based on [this](https://arxiv.org/abs/2302.04222) paper
diff --git a/2023/2023_12_18_glaze_protecting_artists_from_style_mimicry/glaze.pdf b/2023/2023_12_18_glaze_protecting_artists_from_style_mimicry/glaze.pdf
diff --git a/README.md b/README.md
@@ -20,7 +20,7 @@ Join us at https://meet.drwhy.ai.
 * 27.11.2023 - [Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned](https://github.com/MI2DataLab/MI2DataLab_Seminarium/tree/master/2023/2023_11_27_Red_Teaming_Language_Models_to_Reduce_Harms) - Vladimir Zaigrajew
 * 04.12.2023 - Discussion - RedTeaming of foundation models
 * 11.12.2023 - [Introduction to Diffusion Models](https://github.com/MI2DataLab/MI2DataLab_Seminarium/tree/master/2023/2023_12_11_intro_to_diffusion_models) - Bartek Sobieski
-* 18.12.2023 - Glaze: Protecting artists from style mimicry by text-to-image model - Tymoteusz Kwieciński
+* 18.12.2023 - [Glaze: Protecting artists from style mimicry by text-to-image model](https://github.com/MI2DataLab/MI2DataLab_Seminarium/tree/master/2023/2023_12_18_glaze_protecting_artists_from_style_mimicry) - Tymoteusz Kwieciński
 * 08.01.2023 - Red-Teaming the Stable Diffusion Safety Filter - Mateusz Grzyb
 * 15.01.2024 - TBD - Hubert Ruczyński
 * 22.01.2024 - Discussion - Diffusion models for XAI